ID

path_to_ncdf
## [1] "./DataBaseOllieNew/ID116/Trip_EK3/netCDF/"

Pressure

Some context for each track's pressure data

  • Source: Ollie's notebooks and email exchanges between Jelle, Ollie and me

  • The data are not calibrated yet.
  • Since there should be more or less 5-min segments of data every hour, the raw data at 10 Hz are first divided into 5 minute segments (or 15, but it will only contain 5)
  • Data cleaning is done on these segments. So, each segments:
  • Assuming that barometric pressure distributes normally --Jelle said it's a safe assumption --, outliers are filtered out by first obtaining z-scores (gives you an idea of how far from the mean a data point is. But more technically it's a measure of how many standard deviations below or above the population mean a raw score is) and keeping the observations whose absolute z-scores are below 3.
  • Then, if consecutive observations show a difference greater than 2hPa (meaning a ~20m difference), the following value is replaced by the preceding value.
  • Resampling is done at 1 Hz. It is basically done by fitting sinusoids to the data (Fast Fourier Transform) and then taking value points. We're using the obspy.resample function in python which is a wrapper of scipy.signal.resample

  • Everything is saved into a netCDF file

Time series

##   median_BaromP mean_BaromP min_BaromP max_BaromP first_Qu_BaromP
## 1      1027.046    1025.728   1013.744   1031.151        1023.489
##   third_Qu_BaromP mean_IBI_sec mean_samples_within_bout tot_group_n
## 1         1028.47     26368.87                 596.9697          66

Variability within each 5 mins bout in BP

black point: mean blueish bars for min and max greenish bars for q1 and q3 greenish dash lines link medians

Now without max and min

black points: mean squares medians greenish bars for q1 and q3 greenish dash lines link medians

Data to check

Here is plotted how data should look like in first plot

Second plot shows the data within each bout for which the median and the mean differs by more than 0.3hPa.

Infrasound

Some context for each track's infrasound data

  • Source: Ollie's notebooks, email exchanges between Jelle, Ollie and me, some reading and youtube videos

  • The data are calibrated.
  • Since there should be more or less 5-min segments of data every hour (could be 1), the raw data at 10 Hz are first divided into 5 minute segments (or 15, but it will only contain 5)
  • Data cleaning is done on these segments. So, each segments:
  • Assuming that absolute pressure distributes normally, outliers are filtered out by first obtaining z-scores and keeping the observations whose absolute z-scores are below 3.
  • Then, if consecutive observations show a difference greater than 10 (I don't know why 10), the following value is replaced by the preceding value.
  • No resampling
  • For each segment, the power spectral density is estimated using Welch's method; it basically divides data into overlapping segments, computing a modified periodogram for each segment and averaging the periodograms
  • The PSD data are transformed to obtain SPL as \(10 \times \log(PSD/(2*10^{-5})*2))\)

  • The SPL data are saved into a netCDF file

Time series of Sound Pressure Level from Low Infrasound (0.07-1Hz)

##   tot_group_n mean_IBI
## 1          63 27895.74

Time series of Sound Pressure Level from High Infrasound (1-4Hz)